[Bug] Backlog Storage ain't reduced after delete inactive subscription

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/pulsar/issues) and found nothing similar.

### Version

Apache Pulsar: 2.9.3
AKS - K8s version: 1.22.11
Node OS: Ubuntu 18.04

### Minimal reproduce step

1. Create a topic with 2 subscriptions **(sub1, sub2)**
2. Policy configuration: **Retention**: disabled (0, 0), **backlog quota**: size 500GB - time -1, **messageTTL**: 2days
3. Produce messages to that topic
4. Create consumer on **sub1** and consume
5. Wait until backlog size large enough -> Delete subscription **sub2**

### What did you expect to see?

Backlog Storage Size should be reduced

### What did you see instead?

Backlog Storage size isn't reduced, and keep increasing.
On **Local Backlog** chart, blue line corresponds to **sub1**, yellow line is **sub2**

### Anything else?

I get into this issue sometimes while this scenario works normally before, I've no ide. Can someone please point me out what should I do? Is there anyway to delete those storage

It's a big problem for me if data ain't released after cleanup inactive subscriptions.

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Reproduce the listed two-subscription scenario on Apache Pulsar 2.9.3, including disabled retention, a 500GB backlog quota, and subscription deletion. Compare the backlog storage and Local Backlog chart before and after deleting sub2; done means storage is released and no longer continues increasing after cleanup.

Written by the indexing model from the issue text.